Microarrays: Migrating from Screening to Diagnostics

The use of microarrays was first described in 1987 when few cDNAs spotted onto a filter membrane, using a pin spotting device, were used for gene expression profiling. In just a couple of decades the technology rapidly evolved to encompass thousands of genes and sometimes the entire genome of an organism, on a single array. Microarrays are used for studying DNA methylation, copy number changes, gene splicing, microRNA expression, for location analysis to look at protein binding sites on the DNA and for many other biological processes.

"Traditionally this technology has been used for a lot of basic research applications but now you are starting to see pharma and biotech companies using this technology in their actual product development process, looking at toxicological response or in clinical trials for patient stratification and ultimately this technology will work its way into diagnostics," says Kevin Meldrum, senior director of business development for the Genomics division at Agilent Technologies.

Exploring Genetic Diversity

Comparative genomic hybridization (CGH) arrays are an example where the application of the array has evolved over time. CGH arrays were first developed for cytogenetic analysis looking for genetic abnormalities associated with diseases by comparing copy number changes between samples. Paul D. Kassner, PhD, principal scientist at Amgen Inc. started using CGH arrays nearly five years ago looking at amplification and deletion of genes in tumor versus healthy tissue to find out more about the regions in the genome that were likely to function as oncogenes or tumor suppressors. It was later found that even normal, healthy individuals have differences in gene copy number leading to copy number variations (CNV) and these are now considered to be an important source of genetic diversity. "Copy number variations are as important as the single nucleotide polymorphisms (SNPs), which researchers have been profiling for years," says Kassner.

More recently Kassner's group has been using CGH arrays as a target discovery tool in mouse models for detecting copy number variations that could predispose mice to certain diseases. In a study published in the December 2007 issue of Genome Research his group profiled nearly 40 mouse strains, identifying regions that are likely to be deleted and then correlated the copy number changes with certain phenotypes. The phenotype that they studied was food intake, which correlated well with copy number changes in the gene associated with the glucagon-like peptide (GLP-1). CGH arrays are also being used by others as diagnostics for identifying biomarkers for disease pre-disposition and for monitoring patient outcome to drug treatments.

Another growing use for microarrays is in epigenetics. Epigenetics is the study of heritable changes in gene function that cannot be explained by changes in the DNA sequence. Epigenetics plays a role in many biological processes, including cancer and immune response. DNA methylation and histone deactylation are the two most commonly studied epigenetic changes. Microarray-based protocols are now being developed to look at and compare DNA methylation patterns in different cell types or in healthy versus diseased cells, for studying epigenetic changes across the genome, specifically in areas like cancer. "There is a lot of interest in the DNA methylation area but there are not many reliable research tools at this point and a lot of them are low-throughput," says Carson Rosenow, PhD, senior marketing manager for the DNA analysis platform at Illumina Inc. Illumina has launched an array-based high-throughput platform for studying DNA methylation that offers single CpG site resolution.

Screening to Diagnostics

Besides being used effectively as a screening tool, microarrays are also being explored as diagnostics. Bruce J. Paster, PhD, head of the department of Molecular Genetics and professor of Oral Medicine at the Forsyth Institute, a part of the Harvard School of Dental Medicine, is using microarrays to simultaneously determine the profiles of 300 of the most prevalent species of bacteria that colonize the human mouth. "Collectively there are 600 different species of bacteria colonizing the human oral cavity and in any given mouth there are about 75 to100 different species," says Paster. He is planning to use microarray- generated genetic profiles as a tool to distinguish healthy versus disease states, and thus, identify a patient's risk for periodontal disease.

Paster and his colleagues have determined the phylogenetic similarities between different species of bacteria using the 16S ribosomal RNA and used the differences to generate probes for creating oligonucleotide-based reverse capture microarrays, called the Human Oral Microbe Identification Microarrays. "We can grow only about half the species of bacteria that are in the mouth and by using this technique we have generated probes so that we can target the species that we cannot grow," he says.

In January 2008, the Forsyth Microbial Identification Microarray Service for detecting bacterial profiles from oral clinical samples was launched. This service, available worldwide, enables people to send samples to be analyzed on these specifically designed arrays. Paster hopes that in the future the molecular signatures generated by the arrays could be used to tailor treatment options for periodontal disease. Besides targeting the oral cavity, he is also hoping to generate specialized arrays for the bacteria seen in the gastro-intestinal and the urino-gential tract.

While DNA microarrays were quick to be commercialized protein arrays have met with significant challenges. The challenge has been the antibodies. Antibodies are susceptible to denaturing and maintaining the stability of the array over any length of time has been difficult. "Initially it was difficult to find antibodies that could take the abuse of sitting on a surface," says Carl Borrebaeck, PhD, professor and chairman of the department of Immunotechnology at Lund University in Sweden, who has spent many years in antibody engineering. Borrebaeck's group has now generated recombinant antibodies that can remain stable for up to a year. "You can tailormake them so that they are very similar in their physical properties and work almost like a chemical, making them easy to couple or adsorb onto a surface," he says.

Borrebaeck builds and designs his own antibody arrays. "There are no commercially available antibody arrays of significant quality and content for studying the proteins and analytes that are involved in immune response," says Borrebaeck. He has been using antibody arrays to find biomarker signatures from plasma and serum proteins that are predictive for certain types of cancer. In a recent study his group found nine individual analytes that are predictive of metastatic breast cancer in post-menopausal women. In another study, soon to be published, they have identified biomarker signatures in pancreatic carcinomas that can predict patient survival, which in turn can help tailor the drug treatment. "If you are not using a phage display or recombinant antibody library then you have to be very careful and validate how the natural antibody behaves on the surface," says Borrebaeck. "Just because they behave well in other applications does not mean that they will behave well on an array."

While Borrebaeck finds limited use for commercial arrays, Zhiyuan Hu, PhD, research scientist at the Institute for Systems Biology located in Seattle, Washington has met with considerable success testing a commercial protein array instrument being launched in May 2008 by Plexera Biosciences. The instrument, called the Kx5 system, uses surface plasmon resonance (SPR), which is a label-free, high-throughput method for measuring molecular bending events. The array can measure up to 5000 molecular interactions simultaneously in a single run, while providing information about the extent and speed of the bending. "It is going beyond the capabilities of current protein array technologies which usually use fluorescence," says Hu.

Hu has been using this technology for the past year and half to find biomarkers for liver injury. His group has analyzed the human and mouse transcriptome to identify nearly 160 protein targets specific to liver injury. "We have over 100 antibodies to liver specific proteins," says Hu. "Already we have found 10 liver specific proteins changed in the mouse serum and at least five of them are novel biomarkers." The work is still in progress and nearly half the targets still remain to be analyzed. He is also validating the biomarkers in patient samples. Hu hopes that someday this technology too will evolve from a screening to a diagnostic tool.

"The idea is to be able to spot several hundred antibodies onto the chip and then examine the changes using just a drop of blood," he says.

Custom arrays

Microarrays have met with considerable success when used both as a high-throughput, large-scale, screening platform or as a low-throughput, targeted screening tool for answering specific biological questions. "The whole-genome association studies that have been published in the last one and half years have been very exciting, since researchers can now analyze a large sample size in a cost efficient manner to study common diseases," says Rosenow at Illumina, referring to the large collaborative studies that have been published in the areas of diabetes and prostate cancer, an example of the successful high-throughput use of microarrays. At the same time, vendors of the microarray technology are also recognizing the increasing demand for custom arrays geared towards very smallscale, specific applications. "We have seen a tremendous increase in our custom business in the past year and half and we expect that to increase," says Meldrum. To serve those customers, Agilent has launched eArray, an online array creation service, available world-wide, that allows scientists to custom design and print microarrays for their specific application.

However, irrespective of the application, probably one of the biggest limitations of microarrays is that the arrays are designed around reference genomes. For instance, the genome for the commonly used mouse reference strain, C57Black6J, has regions deleted that are present in the other mouse genomes. Hence, there are no probes on the array designed around those deleted regions and hence, information around those regions will be missing from the dataset. "We probably still don't know the areas of the genome that are present or deleted and this is definitely true for the mouse, probably less for the human genome," says Kassner. The hope is that as genomic sequencing becomes more accurate and comprehensive, so will the capabilities of the subsequent microarrays.